摘要
推荐系统是一种解决信息过载的新型技术,为了解决推荐系统中新用户带来的冷启动问题,提出一种基于主动学习的推荐系统。主动学习方法能有效减少需要标记的样本数量,快速建立模型,在此选择将主动学习方法和Baseline SVD推荐算法结合起来,通过记录模型训练得到的预估评价的改变程度,认为改变最大的样例即是最具有信息量的样例,供新用户标记,并重新训练模型。通过与其他选择策略进行实验比较,证实了该方法确实有效解决了新用户带来的冷启动问题。
Recommender system is a new technology to deal with information overload. In order to solve the cold start problem in recommender system,which is brought about by new users,a recommender system based on active learning is presented in this paper. The active learning method can create the model quickly because it can effectively reduce the quantity of training samples needing to be marked. The combination of the active learning method and Baseline SVD recommendation algorithm is adopted in this paper. The change of estimation evaluation is obtained by recording model’s training. The sample which changes most is regarded as the one which is the most informative. Compared with other instance selection strategies,experimental results show that the method can accelerate the speed of cold start brought about by new users.
出处
《现代电子技术》
北大核心
2015年第12期8-11,共4页
Modern Electronics Technique
基金
江苏省社会安全图像与视频理解重点实验室(南京理工大学)开放基金项目(20920130122006)
高等学校学科创新引智计划资助(B13022)